Re: Named RDF Graphs - A Warning

Pardon me for being naive, but here's how I had thought about this.
1. A paper was published in 2004 entitled "Named Graphs, Provenance, and
Trust." I just looked at it and it pretty clearly says that the name of the
named graph is supposed to denote (in the RDF sense) the named graph. Named
graphs are defined as formal pairs (name, rdfgraph).

2. Although I'm not deeply familiar with SPARQL, I assumed that its intent
in allowing one to name graphs was that it was supposed to be compatible
with the Named Graphs paper. This would certainly be parsimonious. Named
graphs are not a W3C recommendation but they seem to have a fair amount of
acceptance - e.g. they are used in the recent OAI-ORE draft standard.

3. Web architecture (according to AWWW and the httpRange-14 resolution) says
nothing that would suggest that a named graph is not a so-called
"information resource" or otherwise suitable as something that has a
so-called "representation" that can be transmitted in an HTTP response or
other message. The reason I say that is that the definition of "information
resource" is so vague and nonactionable that it is a challenge to argue
definitively against any particular thing, and in particular any formally
defined object, as being an "information resource".   Indeed the definition
seems crafted to allow such abstract things to be "information resources".

Together, these considerations would make it very odd indeed to hold a
position other than that a (named) graph (in both the Carroll et al sense
and the SPARQL sense) is an information resource that can have a
representation (i.e. HTTP response) that is an N3 or RDF/XML document. For
named graphs such a "representation" is a bit lossy in that it doesn't carry
the name of the named graph. The architecture takes no explicit stance on
such information loss in going from "essence" to message, and while it is
easy to read AWWW as saying that "representations" must be lossless, no one
seems to do so.

So what you say may be technically true - the URI does not *necessarily*
refer to the (named) graph. However, the analogy with XML namespaces is not
complete because named graphs and SPARQL were both conceived in the context
of RDF, so the expectation that they play nicely with the spirit of RDF and
perhaps even of web architecture is entirely reasonable.

I wouldn't say this is an architecture I particularly like or don't like,
but that's beside the point. I just don't see how given this background
anyone, the TAG or otherwise, is going to prevent a lot of people from
writing RDF statements that use the (named) graph's URI to RDF-refer to the
(named) graph. It's hard enough to get people to recognize that people
aren't documents.

Best
Jonathan


On Dec 18, 2007 7:43 AM, Sean B. Palmer <sean@miscoranda.com> wrote:

>
> Jeremy has proposed @rdf:graph. I like the idea of letting RDF/XML
> talk about graphs, but I worry that we will end up with confusions
> that we had with XML namespaces only worse.
>
> When SPARQL talks about URIs for naming graphs, it doesn't mean that
> that URI identifies the graph. So for example, if you have a graph
> named "http://example.org/powder" that *doesn't* mean that you can
> say:
>
> <http://example.org/powder> a RDFGraph .
>
> This is just like XML namespaces. When you have
> xmlns="http://www.w3.org/1999/xhtml", is the following true?
>
> <http://www.w3.org/1999/xhtml> a Namespace .
>
> No, it is not. The URI is just being used as a unique name. But it's
> very difficult for people to make a new space of URI usage in their
> heads without getting confused, so we end up with this:
>
> "Another benefit of using URIs to build XML namespaces is that the
> namespace URI can be used to identify an information resource that
> contains useful information, machine-usable and/or human-usable, about
> terms in the namespace. This type of information resource is called a
> namespace document."
> - http://www.w3.org/TR/webarch/#namespace-document
>
> This isn't so bad with namespaces, but I think it will be for graphs.
> People will start using using HTTP URIs that return a 200 to identify
> RDF Graphs, whereas RDF Graphs are disjoint with information
> resources. Think about what else you can get from an RDF/XML document:
> an XML infoset, a sequence of unicode characters. The graph is not the
> document; it's not the information resource.
>
> It would be nice to have a URI that identifies a graph, rather than
> just sparql:names it, because it would be less confusing having yet
> another new space for URI usage. I'll bet many people hadn't even
> noticed the difference, and probably still won't understand it after
> reading this message.
>
> Think about it, though. You don't want the TAG kicking down your door at
> 3am.
>
> --
> Sean B. Palmer, http://inamidst.com/sbp/
>
>